literature survey on impulse noise.pdf

8/14/2019 LITERATURE SURVEY ON IMPULSE NOISE.pdf

1/21

Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.5, October 2013

DOI : 10.5121/sipij.2013.4506 75

() & ()

LITERATURE SURVEY ON IMPULSE NOISE

REDUCTION

Manohar Koli1and S.Balaji

2

1Dept. of Computer Science, Tumkur University Tumkur-572103, Karnataka, India

2Dept. of CSE, City Engineering College Bangalore-560061, Karnataka, India

ABSTRACT

In every image processing algorithm quality of image plays a very vital role because the output of the

algorithm depends on the quality of input image. Hence, several techniques are used for image quality

enhancement and image restoration. Some of them are common techniques applied to all the images

without having prior knowledge of noise and are called image enhancement algorithms. Some of the imageprocessing algorithms use the prior knowledge of the type of noise present in the image and are referred to

as image restoration techniques. Image restoration techniques are also referred to as image de-noising

techniques. In such cases, identified inverse degradation functions are used to restore images. In this

survey, we review several impulse noise removal techniques reported in the literature and identify efficient

implementations. We analyse and compare the performance of different reported impulse noise reduction

techniques with Restored Mean Absolute Error (RMAE) under different noise conditions. Also, we identify

the most efficient impulse noise removing filters. Marking the maximum and minimum performance of

filters helps in designing and comparing the new filters which give better results than the existing filters.

KEYWORDS

Impulse Noise, Image Restoration, Image Enhancement, Image De-Noising, Adaptive Filters.

1.INTRODUCTION

Noise is any unwanted signal present in the original signal. Complete classification of noise

is shown in Figure 1. Noise is broadly classified into two main categories: blur noise and

impulse noise.

Figure 1: Classification of Image Noise Types

Blur noise is a uniform noise which makes the image blur. It modifies all the pixels present

in the image by shifting pixel values towards low or high intensity level of the image(Gaussian blur). Sometimes, the noise blurs the image in a particular direction which is


2/21


76

then called the motion blur. Main causes of blur noise are atmospheric turbulence, missedfocus of camera lens, improper opening and closing of camera shutter and relative motion

between camera and object. Impulse noise is a non-uniformly distributed noise. It modifiesonly select pixels from the image keeping the remaining pixels unchanged.

= (1) (1)Where denotes the noiseless image pixel and the noise substituting for the originalpixel.

(a) Original Image. (b) Original Image With 40%Gaussian Noise.

(c) Original Image With 40%Impulse Noise.

Figure 2: Different Types of Noise

Impulse noise produces dot spots or patches on the image. Impulse noise has the property

of either leaving pixels unmodified with probability ( )or replacing it all together witha probability of p as shown in equation (1). The sources of impulse noise are usually the

result of error in transmission system, faulty sensors present in storage or capturingdevices, and atmospheric or man-made disturbances. Outputs of both blur and impulse

noises are shown in Figure 2.

= 0 255 (2) = 0 255 (3)In this paper, both fixed (salt and pepper) and variable (random valued) impulse noise

models are considered for image de-noising. The Salt and Pepper Impulse Noise (SPIN)assumes a minimum value of 0 and a maximum value of 255 of noise, as shown in equation

(2) and Random Valued Impulse Noise (RVIN) method assumes a noise value between the

minimum value of 0 and the maximum value of 255 of noise, as shown in equation (3).

2.FILTERSImpulse noise reduction algorithms are broadly classified in to two classes: linear and non-

linear algorithms, as shown in Figure 3. In literature, many image de-noising algorithms for

correcting the images corrupted by impulse noise are proposed. In a linear technique, thenoise reduction method is applied linearly to all the pixels in the corrupted image without

checking for the corrupted pixels, whereas in non-linear methods corrupted and non-

corrupted pixels are determined first then the reduction techniques are applied forcorrecting the corrupted pixels only. Linear algorithms are time consuming and also blur


3/21


77

the image and hence non-linear noise reduction algorithms are preferred. A set ofconditions are considered in classifying corrupted and non-corrupted pixels in the non-

linear algorithms.

Figure 3: Classification of Reduction Techniques

2.1 Linear Filters

In linear techniques the noise reduction method is applied linearly to all the pixels in the

corrupted image without checking for the corrupted pixels. Linear filters are simple filterswith low computational complexities. Implementation of linear filters is simple compared

to non-linear filters. Mean and median based filters are considered as most popular non-

linear filters. Theoretically, mean based filter produce low mean square error whereas

median based filter produce good noise tolerance.

2.1.1 Mean Filter (MMF)

In simple mean filters odd length fixed sized scanning window is used to get restored imagefrom the corrupted image. With the help of the scanning window corrupted image pixels are

scanned in horizontal and vertical directions. In each scan test pixel is replaced by the

mean value of the scanning window pixels. For our comparative implementation of the

mean filter a (55)sized scanning window is used. In this filter mean value is calculated bytaking ratio of sum of all pixels present in the scanning window and total number of pixels

as shown in equation (4).

(0,0) = (,)()() (4)2.1.2 Weighted Mean Filter (WMMF)Weighted mean filter is another modified basic mean filter. In this filter weights are

assigned to scanning window pixels. Mean value of the products of pixel values and their

corresponding weights are used as restoration value of the test pixel as shown in equation(5). Based on different parameters such as distance from the window center pixel, direction

and position in ordered statistics, etc. different filters use different ways of weight

calculations. Usually, window center pixel is assigned more weights compared to the other

neighboring pixels of the window. Weighted mean filter is used when more than one

parameter or feature of the window pixels influence the restoration value. For our

comparative implementation of weighted mean filter a

(55) sized window and weight

window shown in Figure 4 is used. Weighted window is designed based on the Euclideandistance of pixels from window center pixel.

(0,0) = (,)(,) (,) (5)


4/21


78

1 1 1 1 1

1 2 2 2 1

1 2 3 2 1

1 2 2 2 1

1 1 1 1 1

Figure 4: Weight Window

2.1.3 Trimmed Mean Filter (WMMF)

Trimmed mean filter is an improved version of the mean filter where test pixel is replaced

by trimmed mean value of the window pixels. To calculate the trimmed mean value of the

pixels, trimming value must be known. All available window pixels are arranged inascending order of their intensity values. From ordered list of pixels first and last pixels are removed and the mean value of the remaining pixels is used as the restorationvalue for the corrupted pixels. The process of removing the first and the last set of pixels

from an ordered list is called trimming. The trimming process increases the accuracy or

sharpness of the calculated mean. Calculate the mean of the remaining ordered pixel listand this calculated mean will be used as a restoration value of the test pixel. For our

comparative implementation of the trimmed mean filter a (55) sized window and = 3are used.

2.1.4 Trimmed Weighted Mean Filter (TWMF)

Trimmed weighted mean filter is a combination of both trimmed and weighted mean filters.

This feature catch the power of trimmed filter sharped mean value and also power ofutilizing more than one feature of window pixels. Combined effect of both filters produce

improved performance. For our comparative implementation of weighted trimmed mean

filter (55) sized window, weight window and = 2are used. In this filter window pixelsare arranged in sorted order and then first

and last

pixel values are removed from

ordered list. For restoration weighted mean of remaining pixels is used.2.1.5 Median Filter (MF)

A median filter is the most popularly used one because of its inherent high fault-tolerance.In this filter window pixels values are sorted in order and then the median value is used as

the test pixel value or restoration value. Mean filter is effective in minimizing the mean

square error value of the estimation. Median filter algorithm produces good visibility in therestored image. Many improved versions of median filters are proposed by adding new

features to the existing filters. For our comparative implementation of median filter a(55)window is used. In this filter all selected window pixels are arranged in sorted orderand the median value is used for restoration as shown in equation (6-9).

If n is odd = () (6)(0,0) = (7)If n is even = () (8)


5/21


79

(0,0) = (9)2.1.6 Weighted Median Filter (WMF)

In weighted median filter, weights are assigned to window pixels. All weighted window

pixels are arranged in order and the weighted median value is considered as a replacementvalue of test pixel. For our comparative implementation of weighted median filter a (55)window and weight as shown in Figure 4 is used.

2.1.7 Trimmed Weighted Median Filter (TWMF)

All window pixels are arranged in order and using trimming value first and last values are removed. For the remaining values, the weighted median filter technique is

applied to get the restoration value. For our comparative implementation a (55) weightwindow as shown in Figure 4 and trimming variable =3 are used.2.2 Non-Linear Filters

In non-linear methods corrupted and non-corrupted pixels are determined first and then the

reduction techniques are applied on the corrupted pixels thus detected. Linear algorithms

are time consuming and also blur the image and hence non-linear noise reductionalgorithms are preferred. A set of conditions are considered in identifying corrupted and

non-corrupted pixels in the non-linear algorithms.

2.2.1 Adaptive Median Filter (AMF)

In the adaptive median filter [1], noisy pixels are replaced by the adaptive median value.

Here, adaptive means suitable for the existing noise. According to AMF, median value mustbe present between minimum and maximum values of pixel list. If calculated median value

is not adaptive median then the correct median value is selected by increasing the windowsize of the scanning window. The AMF consists of two levels. The first level tests for the

presence of residual impulses in the median filter output. If the first level asserts that thereis no impulse in the median filter output, then the second level tests whether the center

pixel itself is corrupted by an impulse or not. If the center pixel is decided as uncorrupted,

then AMF leaves it as it is without filtering. If not, the output of the AMF is replaced bythe median filter output at the first level. On the other hand, if the first level asserts that

there is an impulse in the median filter output, then we simply increase the window size forthe median filter and repeat the first-level test.

2.2.2 Progressive Switching Median Filter (PSMF)

Figure 5: Block Diagram of Progressive Switching Median Filter


6/21


80

Progressive switching median filter [2] progressively or iteratively identifies the noisypixels and replaces them iteratively. For the first iteration, a 3x3 window is used and,

gradually, the window size is increased in each iteration. It is switching in nature becausein the first stage it identifies the noisy pixels and it switches to the second phase of the

algorithm for replacing the noisy pixels. First phase of the algorithm, namely, the iterative

noise detection uses the threshold value to compare with the absolute difference of centerpixel and the adaptive median value. If the difference is greater than the test pixel value the

test pixel is considered as noisy pixel and is replaced by the adaptive median value. Thenoisy pixels processed in the one iteration are used to help the process of the other pixels in

the subsequent iterations. The main advantage of this method is that the impulse pixelslocated in the middle of the large noise blotches can also be properly detected and filtered

and, therefore, better restoration results are expected.

2.2.3 Tri-State Median Filter (TSMF)

Tri-state median filter [3], before applying filtering unconditionally, incorporates the

Standard Median (SM) filter and Center Weighted Median (CWM) filter into the noisedetection framework to determine whether the pixel is corrupted. Noise detection isrealized by an impulse detector, which takes the outputs from the SM and CWM filters and

compares them with the origin or center pixel value in order to make a tri-state decision

(Figure 6). The switching logic is controlled by a threshold T and the output of TSM filteris obtained by equation (10).

= 1 2 < 1 < 2 (10)where and are the outputs of and filters, respectively, and 1 =| |and 2 = | |.

Figure 6: Block Diagram of Tri-State Median Filter

2.2.4 Adaptive Fuzzy Switching Filter (AFSF)

The adaptive fuzzy switching filter [4] is composed of three cascaded subunits. The first

subunit aims at detecting impulse noise by considering grayscale distribution amongneighboring pixels. The second subunit implements the grayscale estimation according to

the information of neighboring pixels. The final subunit suitably modifies the value of this

correction in order to further improve the detail preservation by fuzzy switching.


7/21


81

0 0 0 0 0 0 0 -1 0 0 0 0 0 0 -1 -1 0 0 0 0

0 0 0 0 0 0 0 -1 0 0 0 0 0 -1 0 0 -1 0 0 0

-1 -1 4 -1 -1 0 0 4 0 0 0 0 4 0 0 0 0 4 0 0

0 0 0 0 0 0 0 -1 0 0 0 -1 0 0 0 0 0 0 -1 0

0 0 0 0 0 0 0 -1 0 0 -1 0 0 0 0 0 0 0 0 -1

Figure 7: Four 5x5 Convolution Kernels

(,)={ |( , ) |:=14} (11)where is a kernel={(.)} (12) = [(,)(,)(,)(,)] (13)

(, ) = > 0 = 0 (14)

where (.)denotes the pixels not discarded, K denotes the number of pixels having valuenot equal to maximum or minimum value of pixels in the window.

[(,)] [0,1] is the membership function of (,)that indicates how much a pixellooks like an impulse noise. This gives the following fuzzy rules:

[ 1] (, ) , [(, )] . (15)[ 2] (, ) , [(, )] . (16)According to the above rules, s-function is used to describe the membership function of

the impulse noise corruption extent of the current pixel.

[(,)] =0 (, ) < 2 (,) ( , ) 1 2 (,) (, ) 1 (, ) >

(17)where = () Hence, the filter based on adaptive fuzzy rules proposed produces the output value:

(,) = (,) + [(,)] [(,) (,)] (18)

Membership function [(,)]=0means that the current pixel (,)is a noise-free pixeland it needs no filtering. The filter will output the original pixel and preserve the imagedetail. If the membership function [(,)]=1, then the current pixel (,) has beencorrupted absolutely by impulse noise and it needs filtering. The filter will output (,)that is the estimation of the current pixel (,). If the membership function is such that0 < [(,)] < 1, then the current pixel has been corrupted somewhat by impulse noise.The filter will output the weighted average of (,) and (,). Figure 7 shows fourconvolution kernels used in our implementation and analysis.


8/21


82

2.2.5 Novel Impulse Noise Detection (NIND)

The novel impulse noise-detection algorithm [5] is based on the order statistics within a

local window. Impulse detection algorithm is developed based on the following two

assumptions. First, a noise-free image should be locally smoothly varying and is separatedby edges. Second, a noise pixel takes a gray value substantially larger than or smaller than

those of its neighbors. denote the corrupted, noisy image of size 1 2, and is itspixel value at position (i, j ). In order to judge whether is an impulse pixel, first sort the9 pixel values in the 3x3 window centered about it in ascending order. Suppose that the 9

sorted pixel values are (1)


9/21


83

= 0 = 1( )( ) 1 < < 21 >= 2 (21)Where

1 and

2are two pre-determined parameters.

= 0 indicates that the pixel

is noise-free while = 1indicates that it is completely corrupted. When 0 < < 1,it measures how much the pixel is damaged. Calculate the membership function foreach pixel (,); the pixel value of is replaced by a linear combination of its originalvalue and the median ( ) of the local window () with the current pixel excluded.

= (1 ) + ( ) (22)where is the restored value of . It is clear that for a noise-free pixel ( =0),its

value is unchanged, i.e., = . For a heavily corrupted pixel ( = 1), its value isreplaced by the median, i.e., = ( ). For all other pixels (0


10/21


84

Simulation results indicate that this filter is better able to preserve 2-D edge structures ofthe image and delivers better performance with less computational complexity as compared

to other de-noising algorithms existing in literature.

2.2.9 Robust Statistics Based Algorithm (RSBA)

As in [9], let denote the noise corrupted image and for each pixel (, )denoted as , asliding or filtering window of size (2 + 1) (2 + 1)centered at , is defined as shownin Figure 10. The elements of this window are , = { , ,


11/21


85

Algorithm 1

1. Set the minimum window size = 3.2. Read the pixels from the sliding window and store it in .3. Compute minimum (),maximum ()and median value ()inside the window.4. If the center pixel in the window

( , ),is such that


12/21


86

3. The value is checked wirth the predefined threshold value. Based on the conditioncurrent pixel is detected as corrupted or uncorrupted pixel. For optimal performance threshold

value is set as 40 for all images. The above steps are repeated for entire image and binary image of size is obtained using equation (30).

= 1 (, ) < 400 (, ) 40

(30)

Noise Filtering1. A correction window centered at (,) is applied to the flag image , where =2 + 1 and initial is 1. If ( , ) = 0 a result image is obtained by multiplying binary

image segment with noisy image segment by using equation (31). = , , , , (31)2. In , number of non-zero pixels are three and above, (, )is replaced by median of non-zero

pixels in . If numbers of non-zero pixels are less than three, is incremented by one andstep1 and step2 are repeated. The same process is repeated for entire image.

2.2.11 Min-Max Detector Filter (MDBF)[11]

In [11] the authors propose a min-max based simple and efficient algorithm to detect theimpulse noise. A 3x3 scanning window is used to scan the given corrupted image and

center pixel of window is compared with its 8 neighboring pixels present within thewindow. If test pixel is less than the minimum value present in its neighboring pixels or

greater than maximum value then test pixel is considered as corrupted pixel and its value isreplaced by median value of its neighboring pixels. Otherwise, test pixel is considered asnon-corrupted pixel and it is kept as it is and the window is moved to the next pixel. The

algorithm is outlined below.

MDB Filter Algorithm1. Take corrupted image (x).2. Take a 3x3 window (w). Let the center pixel be the test pixel.3. Shift the window row wise then column wise to cover the entire pixel in the image and repeat

step4 and step5.4. If (test pixel< min (rest of the pixel in w) or ( test pixel> max ( rest of the pixel in w) then the

test pixel is corrupted.

5. If the test pixel is corrupted apply median filter to the test filter in the window w.6. Stop.

2.2.12 Detail Preserving Adaptive Filter (DPAF)

The paper [12] proposes an effective and efficient method of impulse noise removal whichnot only removes noise but also preserves the image details. The algorithm first classifies

the pixels as noisy and noise-free based on its N8(p) (Figure 12) neighbours usingaveraging parameters introduced here and then replaces the noisy pixel by the adaptive

median of the pixel. The algorithm uses adaptive median as it provides better de-noising.

Since the proposed algorithm performs prior classification of pixels as noise and noise-free

this method preserves image details. The algorithm is outlined below.

X(i1, j1) X(i1, j) X(i1, j+1)X(i,j1) X(i,j) X(i,j+1)X(i+1, j1) X(i+1, j) X(i+1, j+1)Figure 12: N8(p) of A Pixel


13/21


87

= (32)where = , ,and = , , = (33)where

= , ,and

= , ,

= (34)where = , ,and = , , = (35)where = , ,and = , , = { , , , } (36)

Algorithm1. Input noisy image.2. For all pixels in the image checkXij s if min [w] < < [ w]Xij nc if min[w] = xij or xij = max [w]3. For all pixels in

nccalculate

ah ,av , amd , aadand

and check

If 0.90 1 then xij n,Xij s otherwise4. For all pixels in ,Calculate m. replace each pixe l in n by its m.2.2.13 Universal De-Noising Framework (UDF)

In the filter reported in [13], new detector based on new statistics called RobustOutlyingness Ratio (ROR) is used to measure how much a pixel looks like an impulse

noise. Based on the ROR, the pixels are divided into four different clusters. Then, differentdecision rules are adopted to detect the impulse noise in each cluster. During the detection

process, the from-coarse-to-fine strategy is used. In addition, the detection process containstwo stages, i.e., the coarse stage followed by the fine stage. Different thresholds are used in

the two stages. Finally, the detection procedure is iteratively adopted. In order to calculate

the ROR, the window size must be given. In this paper, the authors use a window size of

5x5. Since the ROR measures the outlyingness of the pixels, i.e., how impulse like, allpixels are divided into four levels (clusters) according to the ROR. The four clusters are:

the most likely cluster ROR; the second likely cluster ROR; the third likely cluster ROR;

and the fourth likely cluster ROR. The lower the ROR, the lower impulse like of the pixelin its neighbors.

The Coarse Stage:1. Choose the algorithm parameters, i.e., coarse thresholds tc1, tc2 , and tc4; window size n

(the actual size is (2n+1)x(2n+1)); iterations mc=1 ; and initial j=1.

2. Initialize the detection flag matrix map as zeros, where 0s and 1s represent good andnoisy pixels, respectively.

3. Calculate the ROR of the current pixel. If the ROR is in the fourth level, treat it as a goodpixel, or calculate the absolute deviation d between the current pixel and the median of its

local window. Then, compare d with tck threshold according to its ROR value. If d is largerthan tck , it is a noisy pixel, or it is a good pixel. Update the flag map according to the

result.

4. Get the median-based restored image i according to the detection result. If the flag is 1,represent the pixel with the median of its local window, or do not change.

5. if j


14/21


88

The Fine Stage:1. Choose the algorithm parameters, i.e., fine thresholds tf1, tf2, tf3, and tf4; window size n

(the actual size is (2n+1)x(2n+1)); iterations mf ; and initial j=1.

2. Initialize the detection flag matrix map as zeros, where 0s and 1s represent good andnoisy pixels, respectively.

3. Calculate the ROR of the current pixel and the absolute deviation between the current pixeland the median of its local window. Then, compare d with threshold tfk according to itsROR value. If is d larger than tfk, it is a noisy pixel, or it is a good pixel. Update the map

flag according to the result.

4. Get the median-based restored image with i the detection result. If the flag is 1, representthe pixel with the median of its local window, or do not change.

5. If j


15/21


89

values of different filters for RVIN images 3 and the outputs are shown in Figures 14; theresults are graphically represented in Figures 16.

TABLE 1: RMAE OF FILTERS FOR SPIN IMAGE-2 (300X300)

NOISE RATIO

10 20 30 40 50 60 70 80 90FILTERS MMF (5X5) -19.8 13.5 25.15 30.04 32.93 34.78 35.15 35.88 36.96

WMMF(5X5) -14.56 15.65 26.04 30.16 33 34.14 35.53 36.38 36.86

TMMF(5X5) 28.2 44.33 41.9 39.25 37.65 35.82 35.35 34.03 33.95

WTMMF(5X5) 29.17 47.27 47.87 46.3 44.17 40.98 39.3 37.47 36.04

MF(5X5) 52.06 73.42 78.91 77.2 64.1 39.87 9.66 -6.56 -6.62

WMF(5X5) 21.2 42.06 42.04 38.42 29.41 13.41 -3.27 -11.28 -7.43

WTMF(5X5) 50.4 73.33 79.82 78.96 68.82 44.96 17.38 -0.46 -4.33

AMF 89.29 92.3 92.69 92.12 90.7 83.87 56.48 28.7 9.82

PSMF 74.36 81.68 83.38 79.97 66.11 35.55 7.34 -3.01 -3.94

TSMF 24.34 61.28 72.35 73.76 64.11 40.34 13.16 -3.69 -4.52

AFSF 77.21 85.48 87.89 87.89 85.39 80.83 72.33 62.05 47.39

NIND 80.87 86.98 86.8 78.08 51.02 13.83 -11.44 -15.78 -8.75

AEAFRIN 72.31 79.39 75.1 63.69 47.89 30.48 15.36 4.86 -0.12

DBA 95.28 95.06 94.11 93.17 92.19 89.95 85.91 70.64 21.19

IAMF 74.82 80.69 81.66 74.88 43.1 15.05 0.02 -10.97 -9.41

RSBA 89.33 91.24 90.87 89.62 88.2 84.41 57.55 23.38 6.63

DBAF 77.62 80.03 71.82 54.93 33.77 15.61 0.1 -7.65 -6.25

MDBF 89.29 90.54 84.94 75.97 61.85 46.27 31.3 18.27 7.99

DPAF 89.41 91.94 92.2 91.51 90.4 83.8 57.34 27.04 9.7

UDF 62.16 78.69 84.01 86.76 87.54 83.15 56.07 18.18 -2.86

TABLE 2: RMAE OF FILTERS FOR RVIN IMAGE-3 (300X300)

NOISE RATIO 10 20 30 40 50 60 70 80 90

FILTERS

MMF (5X5) -80.69 -21.44 -1.46 9.05 14.13 17.45 20 21.47 22.95

WMMF(5X5) -70.68 -15.3 1.64 9.75 14.77 18.34 19.89 21.83 23.19

TMMF(5X5) -29.01 17.48 27.12 28.82 28.48 27.8 26.91 26.49 25.72

WTMMF(5X5) -25.38 18.13 28.09 29.55 28.82 28.78 27.95 27.11 26.33

MF(5X5) 4.6 48.83 62.06 65.68 63.22 54.57 43.75 33.84 26.62

WMF(5X5) -47.3 19.78 40.19 47.76 48.91 41.87 32.87 24.09 16.15

WTMF(5X5) -0.09 46.73 61.2 65.55 63.08 54.98 45.04 35.41 27.85

AMF 56.43 53.92 46.77 39.43 32.32 26.07 20 16.08 12.1

PSMF 49.38 67.3 73.51 74.85 71.49 64.08 51.88 37.45 26

TSMF -34.03 29.27 51 59.05 57.36 50.77 40.51 32.25 25.88

AFSF 40.14 60.61 62.72 58.94 51.68 43.47 34.77 28.03 22.05

NIND 58.86 74.84 79.28 80.83 79.88 74.12 61.15 43.49 23.24

AEAFRIN 39.82 61.8 65.93 62.33 55.95 46.45 37.4 28.54 22.45

DBA 0.76 0.79 0.88 0.91 0.86 0.69 0.8 0.69 0.64

IAMF 52.95 67.3 71.08 71.68 65.6 40.38 17.67 6.94 2.41

RSBA 57.01 53.48 46.57 38.96 32.34 26.37 20.71 15.55 11.93

DBAF 49.44 66.79 68.95 65.04 56.96 47.63 37.63 29.47 22.28

MDBF 57.7 54.9 47.5 39.82 32.85 26.17 20.29 15.65 12.25

DPAF 56.7 54.63 48.18 40.07 32.25 26.3 20.17 16.08 12.4

UDF 17.04 54.92 66.02 70.93 71.57 66.6 54.59 41.67 32.56


16/21

Signal & Image Processin

ORIGINAL IMAGE -2

TMMF(5X5)

WMF(5X5)

PSMF

: An International Journal (SIPIJ) Vol.4, No.5, October

MMF (5X5) WMMF(5

WTMMF(5X5) MF(5X5

WTMF(5X5) AMF

TSMF AFSF

2013

90

X5)

)


17/21

Signal & Image Processin

NIND

IAMF

MDBF

Figure.13 Resul

ORIGINAL IMAGE -3

: An International Journal (SIPIJ) Vol.4, No.5, October

AEAFRIN DBA

RSBA DBAF

DPAF UDF

s Of Filters For Image-2 (300x300) With 60% SPIN

MMF (5X5) WMM

2013

91

(5X5)


18/21


92

TMMF(5X5) WTMMF(5X5) MF(5X5)

WMF(5X5) WTMF(5X5) AMF

PSMF TSMF AFSF

NIND AEAFRIN DBA


19/21


93

IAMF RSBA DBAF

MDBF DPAF UDF

Figure 14: Results of Filters for Image-3 (300x300) with 60% RVIN

Figure 15: Performance of Filters for Image-2 SPIN


20/21


94

Figure 16: Performance of Filters for the Image-3 RVIN

5. CONCLUSIONS

We have implemented most popular and efficient linear and non-linear algorithms for

impulse noise reduction. This paper presents the results and our analysis using differentimages corrupted with SPIN and RVIN. Our experimental results show that the non-linear

filters produce better results compared to the linear filters. Further, our analysis also showsthat handling SPIN is easier than RVIN. For SPIN noise, the DBF, AMF and DPAF filtersshow comparatively better results while for RVIN noise, the NIBOFS, UDF and WMF

filters show comparatively better results.

REFERENCES

[1] H. Hwang and R. A. Haddad Adaptive Median Filters: New Algorithms and Results IEEETransactions on Image Processing, Vol. 4, No. 4, April 1995, pp 499-502.

[2] Zhou Wang and David Zhang Progressive Switching Median Filter for the Removal of Impulse

Noise from Highly Corrupted Images IEEE Transactions on Circuits and SystemsII: Analog and

Digital Signal Processing, Vol. 46, No. 1, January 1999, pp 78-80.

[3] Tao Chen, Kai-Kuang Ma, Li-Hui Chen TriSstate Median Filter for Image Denoising IEEE

Transactions on Image Processing, Vol. 8, No. 12, December 1999, pp 1834-1838.

[4] Haixiang Xu, Guangxi Zhu, Haoyu Peng, Desheng Wang Adaptive Fuzzy Switching Filter for

Images Corrupted by Impulse noise Pattern Recognition Letters 25 (2004) pp 16571663.

[5] Wenbin Luo A New Impulse Detector Based on Order Statistics Intl. J. Electronincs

Communication (ae) 60 (2006) pp 462466.

[6] Wenbin Luo An Efficient Algorithm for the Removal of Impulse Noise from Corrupted Images

Intl. J. Electron. Commun. (ae) 61 (2007) pp 551 555.

[7] K. S. Srinivasan, D. Ebenezer A New Fast and Efficient Decision-Based Algorithm for Removal of

High-Density Impulse Noises IEEE Signal Processing Letters, Vol. 14, No. 3, March 2007, pp 189-192.

[8] Mamta Juneja, Rajni Mohana An Improved Adaptive Median Filtering Method for Impulse Noise

Detection International Journal of Recent Trends in Engineering, Vol. 1, No. 1, May 2009, pp 274-

278.

[9] V.R.Vijaykumar, P.T.Vanathi, P.Kanagasabapathy, D.Ebenezer Robust Statistics Based Algorithm

to Remove Salt and Pepper Noise in Images International Journal of Information and

Communication Engineering 5:3 2009, pp 164-173.


21/21


95

[10] V.R.Vijaykumar, Jothibasu Decision Based Adaptive Median Filter to Remove Blotches, Scratches,

Streaks, Stripes and Impulse Noise in Image Proceedings of 2010 IEEE 17th International

Conference on Image Processing, September 26-29, 2010, Hong Kong, pp 117-120.

[11] S. K. Satpathy, S. Panda, K. K. Nagwanshi, C. Ardil Image Restoration in Non-linear Filtering

Domain Using MDB Approach International Journal of Information and Communication

Engineering 6:1 2010, pp 45-49.

[12] Krishna Kant Singh, Akansha Mehrotra, Kirat Pal, M.J.Nigam A n8(p) Detail Preserving AdaptiveFilter for Impulse Noise Removal 2011 International Conference on Image Information Processing

(ICIIP 2011).[13] Bo Xiong, D. Zhouping Yin A Universal Denoising Framework with a New Impulse Detector and

Non-local Means IEEE Transactions on Image Processing, Vol. 21, No. 4, April 2012, pp 1663-

1675.

Authors Profile

Manohar Koli received a Bachelors degree in Computer Science and Engineering

from Visveswaraya Technological University, Belgaum and a Masters degree in

Computer Science and Engineering from Kuvempu University, Shimoga. He is

currently a research student at Tumkur University, Tumkur and holds the position of

Assistant Professor in Computer Science and Engineering department of University

BDT College of Engineering, Davangere. His research interests include imageprocessing, pattern recognition and data mining.

Balaji S. received a Bachelors degree in Mathematics from Madras University,

Chennai, a Masters degree in Applied Mathematics from Anna University, Chennai

and a Ph.D. degree in Computer Science and Engineering from Indian Institute of

Science, Bangalore. He has served Indian Space Research Organization for more than

two decades before taking up teaching. He is currently a Professor at City Engineering

College, Bangalore. His research interests include fault-tolerant computing, real-time

systems, mobile computing, image processing and computer vision, pattern

recognition, data mining, energy efficient computing, energy management, energy

harvesting, and wireless sensor networks.

literature survey on impulse noise.pdf

Documents